1. Field of the Invention
The invention relates to a high numerical aperture ring field projection optical system for extreme ultraviolet (EUV) photolithography, and particularly to an all-reflective optical system for the camera having six reflecting surfaces with low aspheric departure, and small angles of incidence.
2. Discussion of the Related Art
Short wavelength radiation sources are being used for producing small features on semiconductor wafers, because the size of the smallest features producible using photolithographic techniques, or the critical dimension (CD), depends directly on the wavelength of the source radiation. For this reason, extreme ultraviolet (EUV) radiation is a promising radiation source, having wavelengths in the 4 to 30 nm range. Today, extreme ultraviolet (EUV) projection lithographic technology can be used to produce features sizes that are less than 100 nm.
The photolithographic systems for producing these small feature sizes preferably include all-reflective optics. Dioptric and Catadioptric type projection systems used today for deep ultraviolet (DUV) lithography including refractive optics are not desirable for use in extreme ultraviolet (EUV) lithography systems due to absorption of the EUV radiation in the bulk optical material.
At any given mirror of an optical system, it is useful to quantify the incidence angles of the imaging bundle with respect to the "chief ray." The chief ray from a given field point is the ray that emanates from this field point and passes through the center of the aperture stop. To a good approximation, the mean angle of incidence of any mirror can be estimated by the angle of incidence of the chief ray that emanates from the field point that lies in the center of the ring field. To be more precise, this field point lies in the tangential plane of the projection system at the midpoint of the radial extremum of the arcuate field.
EUV projection lithography is enabled by multilayer coatings that are capable of reflecting approximately 70% of the incident EUV radiation. EUV multilayers include bilayers of Mo/Si and Mo/Be which have been under development for a number of years and are well understood. It is understood in the present invention that since these multilayers rely on a periodic structure to build a reflected wavefront, their performance is sensitive to changes in incidence angle and wavelength.
All-reflective projection lithographic camera systems that support 100 nm resolution with a numerical aperture (NA) in the range of 0.08 to 0.10 are well established in the patent literature. These systems are centered 3- and 4-mirror reflective anastigmats that are optimized to operate over a narrow ringfield. Since it is difficult to control field dependent aberrations (i.e., astigmatism and distortion) to EUV requirements, freedom to control the pupil dependence of the aberrations is necessarily limited. As a result, the numerical aperture of these systems is necessarily restricted to approximately 0.10 for ring field of any substantial width (1.5 mm).
The theoretical resolution (R), or critical dimension (CD) of a lithographic imaging system can be expressed by the well-known relationship R=CD=k.sub.1.lambda./NA, where k.sub.1 is a process dependent constant, .lambda. is the wavelength of light, and NA is the numerical aperture of the projection system. For example, an EUV projection system using a 13.4 nm radiation source and having a k.sub.1 factor of 0.6 and a 0.25 NA can achieve a theoretical resolution on the order of approximately 30 nm. As the critical dimension is reduced, the static distortion should be correspondingly reduced. Specifically, depending on the critical dimension, the static distortion should be reduced to less than about 1/5 at most, and preferably and generally to less than 1/10 of the critical dimension. For example, if CD is 30 nm, then the static distortion should be reduced to less than 3 nm. The static distortion is a measure of the average positional error of focused rays within the focal plane of an optical system. It is desired to have an unobscured multi-mirror projection systems for EUV projection lithography that has both a large numerical aperture that is on the order of 0.25 and low static distortion that is on the order of 1/10 of the critical dimension (CD) or less. Clearly, these projection systems are needed if EUV lithography is to address the sub-100 nm linewidth generations as defined by the SIA roadmap.
One of the first optical systems to address this high NA EUV requirement is disclosed in U.S. Pat No. 5,212,588 entitled, "Reflective Optical Imaging System for Extreme Ultraviolet Wavelengths" issuing to Viswanathan et al., which is herein incorporated by reference. A multi-bounce projection system that incorporates two coaxial aspheric mirrors in a four-bounce configuration is disclosed by Viswanathan and Newnam in the '588 patent and is reproduced at FIG. 1 herein. Mirror M1 is convex and mirror M2 is concave. To obtain high-resolution imagery, the field curvature should be corrected such that the sum of the curvatures of the focusing surfaces of the system, known as the "Petzval sum", are substantially zero. This is known as the "flat field condition." Viswanathan and Newnam designed the system disclosed in the '588 patent so that the two mirrors M1 and M2 have substantially the same radius of curvature, and since M1 is convex and M2 is concave, the flat field condition is satisfied. While the '588 patent describes a number of embodiments with excellent performance over a range of numerical aperture up to 0.3 at DUV wavelengths, all of these embodiments suffer from one common flaw: the exit pupil is centrally obscured by mirror M1. This central obscuration will degrade the MTF response of the system at the mid-spatial frequencies relative to the cut-off frequency. Since the obscuration is large, this loss of contrast will lead to CD variation across the field and yield unacceptable lithographic imaging performance effectively reducing the lithographic process window. Furthermore, the mask (and hence the wafer) plane would be tilted to enable the use of a reflective mask. This introduces added engineering difficulties when a production environment is considered.
The high NA EUV requirement is also discussed briefly by the disclosure in U.S. Pat. No. 5,315,629 entitled, "Ringfield Lithography" issuing to Jewell et al., which is herein incorporated by reference. Jewell and Thompson disclose in the '629 patent that their four-mirror, positive, negative, positive, positive (or PNPP) configuration can be used at a numerical aperture of 0.14, ". . . without significant loss in image quality (still diffraction-limited performance), if the image distortion requirement is relaxed." Jewell and Thompson do not quantify what this relaxation is, but their basic system, reproduced herein at FIG. 2, with a NA of 0.10, has approximately 10 nm of distortion across a 0.5 mm ring. As the numerical aperture of their system is scaled-up, the ability to control the distortion is lost since the available degrees of freedom are consumed to correct aberrations that scale with the NA. A study of the Jewell and Thompson design reveals that the distortion grows to around 30-40 nm at the edge of the ring field as other aberrations are corrected at this increased numerical aperture. This is simply too much distortion for a practical lithographic reduction system, even when the effects of scan-averaging are included.
Neither the two mirror system of Viswanathan nor the four-mirror system of Jewell and Thompson has sufficient degrees of freedom. It is difficult, if not impossible, to configure a four-mirror optical system to produce a system having satisfactorily high NA (i.e., at least 0.25), large ringfield width (i.e., at least 2 mm) and low static distortion (i.e., less than CD/10). For example, starting with either of the systems of FIGS. 1 and 2, the relative positions of the mask, wafer and four mirrors, optimizing the sign and/or magnitude of the curvatures of the four mirrors, and the aspheric profiles of the mirrors, leads to unsatisfactory results.
Ringfield width has a direct influence on wafer throughput, directly influencing the cost of ownership of a lithography system. A system having an increased ringfield width, e.g., 2.0-3.0 mm, has a correspondingly increased throughput. However, the system must still maintain low distortion across its ringfield width if it is to be useful. Static distortion will generally increase with increased ringfield width, but still should be corrected to less than substantially CD/10.
Generally, four-mirror systems simply do not have a sufficient number of degrees of freedom to correct aberrations such as distortion at a numerical aperture of 0.25 over any meaningful field size. The number of degrees of freedom can be increased by adding optical surfaces in such a manner as to enhance the simultaneous correction of both the field and pupil dependent aberrations across the narrow ring field. Since the step and scan method is used to print the entire field, it is practical to investigate solutions with an even number of reflections so that the mask and wafer can be placed on opposing sides of the system to ensure interference-free travel of the mask and wafer. It is desired then to investigate systems having at least six reflective surfaces.
Recently, optical projection reduction systems have been disclosed that offer high numerical apertures with six and eight reflections. One such system is disclosed in U.S. Pat No. 5,686,728 entitled, "Projection Lithography System and Method Using All-Reflective Optical Elements," issuing to Shafer, which is herein incorporated by reference. In the '728 patent, Shafer describes an eight mirror projection system with a numerical aperture of around 0.50 and a six mirror projection system with a numerical aperture of around 0.45. The eight and six mirror systems of Shafer are reproduced, respectively, at FIGS. 3a and 3b herein.
These systems were designed for DUV lithography, and, while fine for that purpose, are not necessarily suitable for EUV projection lithography, even after the NA has been reduced from around 0.50 to around 0.25. By way of example, mirror M1 in the six mirror embodiment of FIG. 3b is essentially flat making it difficult to test with state of the art visible light interferometers designed to measure aspheric mirrors to the required accuracy (see Gary E. Sommergren, "Phase Shifting Diffraction Interferometry for Measuring Extreme Ultraviolet Optics", OSA TOPS on Extreme Ultraviolet Lithography, Vol. 4, pp. 108-112 (1996)).
Another high numerical aperture projection system is the disclosure in U.S. Pat. No. 5,815,310 entitled, "High Numerical Aperture Ring Field Optical Reduction System" issuing to Williamson, which is herein incorporated by reference. In the '310 patent, Williamson describes two six mirror ring field projection systems intended for use with both DUV and EUV radiation. A first arrangement of Williamson, reproduced herein at FIG. 4a, consists, from long conjugate to short conjugate, of concave, convex, concave, concave, convex, and concave surfaces, or PNPPNP for short. This projection system has a numerical aperture of 0.25 and is intended for EUV radiation. A second arrangement of Williamson, reproduced herein at FIG. 4b consists, from long conjugate to short conjugate, of concave, convex, concave, concave, convex, and concave surfaces, or PNPPNP for short. This projection system has a numerical aperture of 0.55 and is intended for DUV radiation. Referring specifically to FIG. 4a, the projection system is capable of 30 nm lithography using conservative values for k.sub.1 around 0.6. Williamson discloses both the PNPPNP and the PPPPNP reimaging configurations locating an intermediate image in various locations, always near the mirror pair comprising the third and fourth concave mirrors.
The EUV arrangement suffers from the following drawbacks. First, large peak aspheric departures are present on each of mirror M1 and mirror M5. To enable tests of the surface figure, a complex null lens or computer generated hologram (CGH) is needed to provide the required aspheric reference wavefront. Using a null lens or CGH would seriously compromise the absolute accuracy of the test and could potentially lead to errors that would prevent convergence to the proper aspheric figure. Aspheric departures on the order of less than approximately 15 .mu.m across any reasonably sized clear aperture would enable a center of curvature test using visible light metrology without using a null lens or CGH. The peak departures on mirrors M1 and M5 of the '310 patent of Williamson are 30 .mu.m and 18.5 .mu.m, respectively, across the clear aperture of the off-axis section. In all likelihood, these large departures require the use of a null lens, a CGH or a data reduction scheme whereby multiple subaperture data sets are "stitched" together. Neither of these metrology techniques is preferred. A system with mirrored surfaces each having low aspheric departure is desired.
Another issue with the systems disclosed by Williamson is the high incidence angles at each of the mirrored surfaces, particularly on mirrors M2 and M3. In some instances, the angle of incidence exceeds 24.degree. at a given location of a mirror. Both the mean angle and deviation or spread of angles is sufficient to cause noticeable amplitude and phase effects due to the EUV multilayer coatings.